I joined the Openclaw crowd a bit late. I will update this post from time to time with notes on Claw configuration and mysterious skills. Fellow Claw users are welcome to discuss.

JaneZ-Claw

Actually, it should probably be JaneA-Claw, since JaneA’s owner is JaneZ.

Setup

Like most people, I used an Alibaba Cloud server. First you need to buy an Alibaba Cloud Lightweight Application Server, set the application image to openclaw, and choose a one-year term. Since I missed the golden window for joining, the price had already increased from 68 RMB/year to 79 RMB/year. I paid 11 RMB extra for nothing. If you want to set one up, do it early, since the price will probably keep rising.

The following steps are very beginner-friendly. Wait 1 to 5 minutes until the server deployment is complete, then connect with one click, choose your API, enter the API key, and you can get the web address. Open it and you will see the xxx-Claw interface.

API Purchases

Recently I bought APIs for several types of models, including text-generation models, image-generation models, and multimodal models. Prices range from tens to hundreds of RMB. I am still unsure which image-generation or multimodal models are actually better. The true god is nano banana pro, but I cannot afford it. A funny side story: when using nano banana pro, the generated images famously come with watermarks, so I found a somewhat cursed method. Install the Nano Banana Watermark Remover Chrome extension and it removes the watermark. It is sketchy, but it works well. Of course, this kind of extension seriously harms Google’s interests, so Chrome will probably ban it soon. For now I will mainly share my experience with several domestic text-generation APIs. I need to find a way to get an overseas Visa card.

Text-Generation Models

Deepseek V3.2

  • base_url: https://api.deepseek.com
  • Purchase method: no package plans, just top up directly. This is my favorite part.
  • Price: I barely did anything and already spent 0.13 RMB.
  • Experience: terrible. If, like me, you like clean and clear payment rules, then feel free to use DeepSeek. Even eating dirt feels voluntary.

Qwen 3.5-plus

  • base_url: https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions
  • Purchase method: Alibaba Cloud’s UI is a mess. I do not even want to talk about it. I searched for ages and could not find where to buy it, nor where to check token usage. It does occasionally send you SMS reminders telling you how many tokens you have left and that you should top up. It is package-based. The original wording is used to offset model inference fees generated after exceeding the free quota. It costs 20 RMB for the first month and 100 RMB for the first three months. There also seems to be an option to top up tokens directly, but the billing rules are complicated.
  • Price: It gave about 200k free tokens. The bot burned around 13k tokens after only two or three messages.
  • Experience: I think I still respect Qwen quite a bit. The small model performs decently.

GLM-5

  • base_url: https://open.bigmodel.cn/api/paas/v4/chat/completions
  • Purchase method: like DeepSeek, top up directly, no package plan, or maybe I just did not see one.
  • Price: real-name verification gives a million tokens, which feels nice.
  • Experience: seems okay and quite fast, but I have not used it much.

A Joke

Image-Generation Models & Multimodal Models

TBD

Bot Settings

Telegram login has never succeeded. I suspect it is a proxy problem. I will update this after I fix it, though I may also never update it, since I have never used Telegram and do not really want to.

QQ Bot: JaneA

I honestly think this thing is not very useful and is rather dumb, but I will still attach the setup tutorial because it is not hard.


In the channel configuration section of your server application details page, enter your QQ bot APP ID and APP secret, then click Apply.

The effect is just so-so. Here are some usage records.

Workflow Session

TBD. I am designing a decent workflow with @Parsifal1986. I will update this after it is ready.

Skill

TBD. I am trying out some skills and will recommend a few.